| Message ID | 20231007203543.1377452-6-willy@infradead.org (mailing list archive) |
|---|---|
| State | Deferred, archived |
| Headers | show |
| Series | Remove the XFS mrlock | expand |
On 10/7/23, Matthew Wilcox (Oracle) <willy@infradead.org> wrote: > +static inline void inode_assert_locked_excl(const struct inode *inode) > +{ > + rwsem_assert_held_write(&inode->i_rwsem); > +} > + > static inline void inode_lock_nested(struct inode *inode, unsigned > subclass) > { > down_write_nested(&inode->i_rwsem, subclass); Why "excl" instead of "write"? Apart from looking weird, it is inconsistent with "prior art" in the file: i_mmap_assert_write_locked.
On Sun, Oct 08, 2023 at 10:26:40PM +0200, Mateusz Guzik wrote: > On 10/7/23, Matthew Wilcox (Oracle) <willy@infradead.org> wrote: > > +static inline void inode_assert_locked_excl(const struct inode *inode) > > +{ > > + rwsem_assert_held_write(&inode->i_rwsem); > > +} > > + > > static inline void inode_lock_nested(struct inode *inode, unsigned > > subclass) > > { > > down_write_nested(&inode->i_rwsem, subclass); > > Why "excl" instead of "write"? Apart from looking weird, it is > inconsistent with "prior art" in the file: i_mmap_assert_write_locked. Yes, but that pairs with i_mmap_lock_write() / i_mmap_lock_read(). The problem is that we have inode_lock() / inode_lock_shared() inode_assert_locked_read/write doesn't make sense with them. But inode_assert_locked() doesn't make sense as the assertion for inode_lock() because you'd expect it to assert whether the inode lock is held at all. So I went with inode_assert_locked_excl(). I wouldn't mind if we converted all the inode_lock()/shared to inode_lock_read() / inode_lock_write(), and then added inode_assert_read_locked() / inode_assert_write_locked(). That's a bit of a bigger job than I want to take on today.
On 10/8/23, Matthew Wilcox <willy@infradead.org> wrote: > On Sun, Oct 08, 2023 at 10:26:40PM +0200, Mateusz Guzik wrote: >> On 10/7/23, Matthew Wilcox (Oracle) <willy@infradead.org> wrote: >> > +static inline void inode_assert_locked_excl(const struct inode *inode) >> > +{ >> > + rwsem_assert_held_write(&inode->i_rwsem); >> > +} >> > + >> > static inline void inode_lock_nested(struct inode *inode, unsigned >> > subclass) >> > { >> > down_write_nested(&inode->i_rwsem, subclass); >> >> Why "excl" instead of "write"? Apart from looking weird, it is >> inconsistent with "prior art" in the file: i_mmap_assert_write_locked. > > Yes, but that pairs with i_mmap_lock_write() / i_mmap_lock_read(). > > The problem is that we have inode_lock() / inode_lock_shared() > inode_assert_locked_read/write doesn't make sense with them. But > inode_assert_locked() doesn't make sense as the assertion for > inode_lock() because you'd expect it to assert whether the inode lock > is held at all. So I went with inode_assert_locked_excl(). > > I wouldn't mind if we converted all the inode_lock()/shared to > inode_lock_read() / inode_lock_write(), and then added > inode_assert_read_locked() / inode_assert_write_locked(). That's > a bit of a bigger job than I want to take on today. > I agree it is rather messy and I'm not going to spend time arguing as it is not my call anyway. Speaking of that, I just noticed the vfs folk are not CC'ed, which I'm rectifying with this e-mail.
diff --git a/fs/attr.c b/fs/attr.c index a8ae5f6d9b16..5e32b0a4f8c2 100644 --- a/fs/attr.c +++ b/fs/attr.c @@ -387,7 +387,7 @@ int notify_change(struct mnt_idmap *idmap, struct dentry *dentry, struct timespec64 now; unsigned int ia_valid = attr->ia_valid; - WARN_ON_ONCE(!inode_is_locked(inode)); + inode_assert_locked_excl(inode); error = may_setattr(idmap, inode, ia_valid); if (error) diff --git a/fs/namei.c b/fs/namei.c index 567ee547492b..6b595ad4318d 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -2708,7 +2708,7 @@ struct dentry *try_lookup_one_len(const char *name, struct dentry *base, int len struct qstr this; int err; - WARN_ON_ONCE(!inode_is_locked(base->d_inode)); + inode_assert_locked(base->d_inode); err = lookup_one_common(&nop_mnt_idmap, name, base, len, &this); if (err) @@ -2735,7 +2735,7 @@ struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) struct qstr this; int err; - WARN_ON_ONCE(!inode_is_locked(base->d_inode)); + inode_assert_locked(base->d_inode); err = lookup_one_common(&nop_mnt_idmap, name, base, len, &this); if (err) @@ -2765,7 +2765,7 @@ struct dentry *lookup_one(struct mnt_idmap *idmap, const char *name, struct qstr this; int err; - WARN_ON_ONCE(!inode_is_locked(base->d_inode)); + inode_assert_locked(base->d_inode); err = lookup_one_common(idmap, name, base, len, &this); if (err) diff --git a/include/linux/fs.h b/include/linux/fs.h index b528f063e8ff..e01e041c102b 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -832,6 +832,16 @@ static inline int inode_is_locked(struct inode *inode) return rwsem_is_locked(&inode->i_rwsem); } +static inline void inode_assert_locked(const struct inode *inode) +{ + rwsem_assert_held(&inode->i_rwsem); +} + +static inline void inode_assert_locked_excl(const struct inode *inode) +{ + rwsem_assert_held_write(&inode->i_rwsem); +} + static inline void inode_lock_nested(struct inode *inode, unsigned subclass) { down_write_nested(&inode->i_rwsem, subclass);
Use the new rwsem_assert_held functions to implement these new assertions. Convert the inode_is_locked() callers in the VFS to use them. Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- fs/attr.c | 2 +- fs/namei.c | 6 +++--- include/linux/fs.h | 10 ++++++++++ 3 files changed, 14 insertions(+), 4 deletions(-)